Hacker News Reader

Comment by eadmund | Hacker News Reader

If there is no attorney-client privilege with AI (I believe there is not), then could the opposing party demand discovery of all of one’s communications with the AI pertaining to the case?

View on HN

Lawyer here. There isn’t attorney-client privilege with AI itself because AI is neither your attorney nor your client. However, there is something called the work product doctrine that shields things like your activities with legal research tools, document review tools, etc. so long as they are used in anticipation of litigation. So, despite what the other commenters are saying here, there is pretty much no chance that any of this would be discoverable, barring very strange circumstances.

I'm curious where the line is if you represent yourself in a criminal case. At what point (if any) would opening up Google Docs and typing notes about the case that inadvertantly incriminates yourself become inadmissible?

I don't have an answer to this- but if the above answer was based on the Federal Rules of Civil Procedure it may not apply to criminal cases at all. (And the above was probably a generalization that assumes parties in California or some other state the lawyer was familiar with.)

Let me preface this by saying that civil procedure and criminal procedure are separate bodies of law with their own rules on these issues. I’m not a criminal attorney and thus not comfortable speaking out of field. That said, I know there is an analogous work product doctrine that applies to pro se criminal defendants, but I don’t know the contours of it.

It’s super interesting to think about though! Imagine if your Google Doc said something like “The stolen cash is hidden in these six places, but the prosecutors don’t seem to have anything linking me to some of those areas, so I should point that out at trial.” If the government didn’t already know where the cash was hidden, it seems implausible to me that any judge wouldn’t let them have that information. It likewise seems implausible that the judge would let the prosecutors know what the defendant thinks would be good or bad to emphasize at trial. So I’m guessing it’d be an in-camera review resulting in a redacted document just disclosing the places where the cash was hidden, but not the subsequent mental impressions. But the real question, as you asked, is whether you can actually put that in front of a jury to show guilt. And I’m afraid I’m going to have to plead ignorance here (but hope there are some criminal attorneys lurking on HN that could speak to this).

That said, it would have to be disclosed via a privilege log. You would probably wind up having to disclose what documents you uploaded. Your prompts and the responses are probably protected work product.

Better advice would cover exactly what's needed to maximize that odds of that determination. For example, commingling the 'research' with activity that is the subject of the ligation may make your life much harder (or at least more expensive) because your opponent is absolutely entitled to that other material at the very least.

It also doesn't help when your opponent hit the provider ex parte with a subpoena and they just served it up which I can speak from experience many providers are pretty eager to do. Maybe after the fact you get an order for your opponent to destroy and not use the material but if you're dealing with a party that was dishonest enough to bring a frivolous claim in the first place that may not be that comforting.

The interesting question - if the AI is provided via my lawyer's office - is there privilege in this case?

There's no privilege, so yes, if the AI chats are retained, they may be subject to discovery.

even for say google search histories?

Theoretically yes, if the court can be convinced that it is relevant to the case.

This is completely wrong, as the other commenter stated. This would certainly fall under the work-product doctrine [1,2], where documents prepared for the purpose of future litigation are protected from discovery and could be considered analogous to attorney-client privilege (but is actually much more broadly defined than attorney-client privilege[4]). Even if opposing counsel is able to obtain discovery on a work-product, only fact based products, not opinion based are allowed. In other words, the court removes anything related to “mental impressions, conclusions, opinions, or legal theories of an attorney or other representative of a party concerning the litigation” [3]. For conversations with AI about how to conduct your case, that would exclude basically everything since it is an opinion work-product, not a fact work product. A fact based work-product would be things like “statements or interviews of now deceased witnesses, photographs or video of an accident scene taken at the time of the accident”[4].

[1] https://www.floridabar.org/the-florida-bar-journal/the-work-...

[2] https://en.wikipedia.org/wiki/Work-product_doctrine

[3] https://coxlawflorida.com/florida-rules-of-civil-procedure/r...

[4] https://tenthings.blog/2019/06/05/ten-things-a-primer-on-the...

Thank you

Reply on news.ycombinator.com